Benchmark on Automatic 6-month-old Infant Brain Segmentation Algorithms: The iSeg-2017 Challenge

A foundation model for enhancing magnetic resonance images and downstream segmentation, registration and diagnostic tasks

In structural magnetic resonance (MR) imaging, motion artefacts, low resolution, imaging noise and variability in acquisition protocols frequently degrade image quality and confound downstream analyses. Here we report a foundation model for the motion correction, resolution enhancement, denoising and harmonization of MR images. Specifically, we trained a tissue-classification neural network to predict tissue labels, which are then leveraged by a 'tissue-aware' enhancement network to generate high-quality MR images. We validated the model's effectiveness on a large and diverse dataset comprising 2,448 deliberately corrupted images and 10,963 images spanning a wide age range (from foetuses to elderly individuals) acquired using a variety of clinical scanners across 19 public datasets. The model consistently outperformed state-of-the-art algorithms in improving the quality of MR images, handling pathological brains with multiple sclerosis or gliomas, generating 7-T-like images from 3 T scans and harmonizing images acquired from different scanners. The high-quality, high-resolution and harmonized images generated by the model can be used to enhance the performance of models for tissue segmentation, registration, diagnosis and other downstream tasks.

Human lifespan changes in the brain's functional connectome

Functional connectivity of the human brain changes through life. Here, we assemble task-free functional and structural magnetic resonance imaging data from 33,250 individuals at 32 weeks of postmenstrual age to 80 years from 132 global sites. We report critical inflection points in the nonlinear growth curves of the global mean and variance of the connectome, peaking in the late fourth and late third decades of life, respectively. After constructing a fine-grained, lifespan-wide suite of system-level brain atlases, we show distinct maturation timelines for functional segregation within different systems. Lifespan growth of regional connectivity is organized along a spatiotemporal cortical axis, transitioning from primary sensorimotor regions to higher-order association regions. These findings elucidate the lifespan evolution of the functional connectome and can serve as a normative reference for quantifying individual variation in development, aging and neuropsychiatric disorders.

Age-specific optimization of the T2-weighted MRI contrast in infant and toddler brain

PURPOSE

In 0-2-year-old brains, the T2-weighted (T2w) contrast between white matter (WM) and gray matter (GM) is weaker compared with that in adult brains and rapidly changes with age. This study aims to design variable-flip-angle (VFA) trains in 3D fast spin-echo sequence that adapt to the dynamically changing relaxation times to improve the contrast in the T2w images of the developing brains.

METHODS

T1 and T2 relaxation times in 0-2-year-old brains were measured, and several age groups were defined according to the age-dependent pattern of T2 values. Based on the static pseudo-steady-state theory and the extended phase graph algorithm, VFA trains were designed for each age group to maximize WM/GM contrast, constrained by the maximum specific absorption rate and overall signal intensity. The optimized VFA trains were compared with the default one used for adult brains based on the relative contrast between WM and GM. Dice coefficient was used to demonstrate the advantage of contrast-improved images as inputs for automatic tissue segmentation in infant brains.

RESULTS

The 0-2-year-old pool was divided into groups of 0-8 months, 8-12 months, and 12-24 months. The optimal VFA trains were tested in each age group in comparison with the default sequence. Quantitative analyses demonstrated improved relative contrasts in infant and toddler brains by 1.5-2.3-fold at different ages. The Dice coefficient for contrast-optimized images was improved compared with default images (p < 0.001).

CONCLUSION

An effective strategy was proposed to improve the 3D T2w contrast in 0-2-year-old brains.

Enhanced U-Net for Infant Brain MRI Segmentation: A (2+1)D Convolutional Approach

BACKGROUND

Infant brain tissue segmentation from MRI data is a critical task in medical imaging, particularly challenging due to the evolving nature of tissue contrasts in the early months of life. The difficulty increases as gray matter (GM) and white matter (WM) intensities converge, making accurate segmentation challenging. This study aims to develop an improved U-net-based model to enhance the precision of automatic segmentation of cerebro-spinal fluid (CSF), GM, and WM in 10 infant brain MRIs using the iSeg-2017 dataset.

METHODS

The proposed method utilizes a U-net architecture with (2+1)Dconvolutional layers and skip connections. Preprocessing includes intensity normalization using histogram alignment to standardize MRI data across different records. The model was trained on the iSeg-2017 dataset, which comprises T1-weighted and T2-weighted MRI data from ten infant subjects. Cross-validation was performed to evaluate the model's segmentation performance.

RESULTS

The model achieved an average accuracy of 92.2%, improving on previous methods by 0.7%. Sensitivity, precision, and Dice similarity scores were used to evaluate the performance, showing high levels of accuracy across different tissue types. The model demonstrated a slight bias toward misclassifying GM and WM, indicating areas for potential improvement.

CONCLUSIONS

The results suggest that the U-net architecture is highly effective in segmenting infant brain tissues from MRI data. Future work will explore enhancements such as attention mechanisms and dual-network processing for further improving segmentation accuracy.

Structure-Aware Brain Tissue Segmentation for Isointense Infant MRI Data Using Multi-Phase Multi-Scale Assistance Network

Accurate and automatic brain tissue segmentation is crucial for tracking brain development and diagnosing brain disorders. However, due to inherently ongoing myelination and maturation during the first postnatal year, the intensity distributions of gray matter and white matter in the infant brain MRI at the age of around 6 months old (a.k.a. isointense phase) are highly overlapped, which makes tissue segmentation very challenging, even for experts. To address this issue, in this study, we propose a multi-phase multi-scale assistance segmentation framework, which comprises a structure-preserved generative adversarial network (SPGAN) and a multi-phase multi-scale assisted segmentation network (MASN). SPGAN bi-directionally synthesizes isointense and adult-like data. The synthetic isointense data essentially augment the training dataset, combined with high-quality annotations transferred from its adult-like counterpart. By contrast, the synthetic adult-like data offers clear tissue structures and is concatenated with isointense data to serve as the input of MASN. In particular, MASN is designed with two-branch networks, which simultaneously segment tissues with two phases (isointense and adult-like) and two scales by also preserving their correspondences. We further propose a boundary refinement module to extract maximum gradients from local feature maps to indicate tissue boundaries, prompting MASN to focus more on boundaries where segmentation errors are prone to occur. Extensive experiments on the National Database for Autism Research and Baby Connectome Project datasets quantitatively and qualitatively demonstrate the superiority of our proposed framework compared with seven state-of-the-art methods.

GLAC-Unet: Global-Local Active Contour Loss with an Efficient U-Shaped Architecture for Multiclass Medical Image Segmentation

The field of medical image segmentation powered by deep learning has recently received substantial attention, with a significant focus on developing novel architectures and designing effective loss functions. Traditional loss functions, such as Dice loss and Cross-Entropy loss, predominantly rely on global metrics to compare predictions with labels. However, these global measures often struggle to address challenges such as occlusion and nonuni-form intensity. To overcome these issues, in this study, we propose a novel loss function, termed Global-Local Active Contour (GLAC) loss, which integrates both global and local image features, reformulated within the Mumford-Shah framework and extended for multiclass segmentation. This approach enables the neural network model to be trained end-to-end while simultaneously segmenting multiple classes. In addition to this, we enhance the U-Net architecture by incorporating Dense Layers, Convolutional Block Attention Modules, and DropBlock. These improvements enable the model to more effectively combine contextual information across layers, capture richer semantic details, and mitigate overfitting, resulting in more precise segmentation outcomes. We validate our proposed method, namely GLAC-Unet, which utilizes the GLAC loss in conjunction with our modified U-shaped architecture, on three biomedical segmentation datasets that span a range of modalities, including two-dimensional and three-dimensional images, such as dermoscopy, cardiac magnetic resonance imaging, and brain magnetic resonance imaging. Extensive experiments demonstrate the promising performance of our approach, achieving a Dice score (DSC) of 0.9125 on the ISIC-2018 dataset, 0.9260 on the Automated Cardiac Diagnosis Challenge (ACDC) 2017, and 0.927 on the Infant Brain MRI Segmentation Challenge 2019. Furthermore, statistical significance testing with p-values consistently smaller than 0.05 on the ISIC-2018 and ACDC datasets confirms the superior performance of the proposed method compared to other state-of-the-art models. These results highlight the robustness and effectiveness of our multiclass segmentation technique, underscoring its potential for biomedical image analysis. Our code will be made available at https://github.com/minhnhattrinh312/Active-Contour-Loss-based-on-Global-and-Local-Intensity.

Convolutional neural network (CNN) configuration using a learning automaton model for neonatal brain image segmentation

CNN is considered an efficient tool in brain image segmentation. However, neonatal brain images require specific methods due to their nature and structural differences from adult brain images. Hence, it is necessary to determine the optimal structure and parameters for these models to achieve the desired results. In this article, an adaptive method for CNN automatic configuration for neonatal brain image segmentation is presented based on the encoder-decoder structure, in which the hyperparameters of this network, i.e., size, length, and width of the filter in each layer along with the type of pooling functions with a reinforcement learning approach and an LA model are determined. These LA models determine the optimal configuration for the CNN model by using DICE and ASD segmentation quality evaluation criteria, so that the segmentation quality can be maximized based on the goal criteria. The effectiveness of the proposed method has been evaluated using a database of infant MRI images and the results have been compared with previous methods. The results show that by using the proposed method, it is possible to segment NBI with higher quality and accuracy.

Brain volume in infants with metopic synostosis: Less white matter volume with an accelerated growth pattern in early life

Metopic synostosis patients are at risk for neurodevelopmental disorders despite a negligible risk of intracranial hypertension. To gain insight into the underlying pathophysiology of metopic synostosis and associated neurodevelopmental disorders, we aimed to investigate brain volumes of non-syndromic metopic synostosis patients using preoperative MRI brain scans. MRI brain scans were processed with HyperDenseNet to calculate total intracranial volume (TIV), total brain volume (TBV), total grey matter volume (TGMV), total white matter volume (TWMV) and total cerebrospinal fluid volume (TCBFV). We compared global brain volumes of patients with controls corrected for age and sex using linear regression. Lobe-specific grey matter volumes were assessed in secondary analyses. We included 45 metopic synostosis patients and 14 controls (median age at MRI 0.56 years [IQR 0.36] and 1.1 years [IQR 0.47], respectively). We found no significant differences in TIV, TBV, TGMV or TCBFV in patients compared to controls. TWMV was significantly smaller in patients (-62,233 mm3 [95% CI = -96,968; -27,498], Holm-corrected p = 0.004), and raw data show an accelerated growth pattern of white matter in metopic synostosis patients. Grey matter volume analyses per lobe indicated increased cingulate (1378 mm3 [95% CI = 402; 2355]) and temporal grey matter (4747 [95% CI = 178; 9317]) volumes in patients compared to controls. To conclude, we found smaller TWMV with an accelerated white matter growth pattern in metopic synostosis patients, similar to white matter growth patterns seen in autism. TIV, TBV, TGMV and TCBFV were comparable in patients and controls. Secondary analyses suggest larger cingulate and temporal lobe volumes. These findings suggest a generalized intrinsic brain anomaly in the pathophysiology of neurodevelopmental disorders associated with metopic synostosis.

A flexible 2.5D medical image segmentation approach with in-slice and cross-slice attention

Deep learning has become the de facto method for medical image segmentation, with 3D segmentation models excelling in capturing complex 3D structures and 2D models offering high computational efficiency. However, segmenting 2.5D images, characterized by high in-plane resolution but lower through-plane resolution, presents significant challenges. While applying 2D models to individual slices of a 2.5D image is feasible, it fails to capture the spatial relationships between slices. On the other hand, 3D models face challenges such as resolution inconsistencies in 2.5D images, along with computational complexity and susceptibility to overfitting when trained with limited data. In this context, 2.5D models, which capture inter-slice correlations using only 2D neural networks, emerge as a promising solution due to their reduced computational demand and simplicity in implementation. In this paper, we introduce CSA-Net, a flexible 2.5D segmentation model capable of processing 2.5D images with an arbitrary number of slices. CSA-Net features an innovative Cross-Slice Attention (CSA) module that effectively captures 3D spatial information by learning long-range dependencies between the center slice (for segmentation) and its neighboring slices. Moreover, CSA-Net utilizes the self-attention mechanism to learn correlations among pixels within the center slice. We evaluated CSA-Net on three 2.5D segmentation tasks: (1) multi-class brain MR image segmentation, (2) binary prostate MR image segmentation, and (3) multi-class prostate MR image segmentation. CSA-Net outperformed leading 2D, 2.5D, and 3D segmentation methods across all three tasks, achieving average Dice coefficients and HD95 values of 0.897 and 1.40 mm for the brain dataset, 0.921 and 1.06 mm for the prostate dataset, and 0.659 and 2.70 mm for the ProstateX dataset, demonstrating its efficacy and superiority. Our code is publicly available at: https://github.com/mirthAI/CSA-Net.

A review of self-supervised, generative, and few-shot deep learning methods for data-limited magnetic resonance imaging segmentation

Magnetic resonance imaging (MRI) is a ubiquitous medical imaging technology with applications in disease diagnostics, intervention, and treatment planning. Accurate MRI segmentation is critical for diagnosing abnormalities, monitoring diseases, and deciding on a course of treatment. With the advent of advanced deep learning frameworks, fully automated and accurate MRI segmentation is advancing. Traditional supervised deep learning techniques have advanced tremendously, reaching clinical-level accuracy in the field of segmentation. However, these algorithms still require a large amount of annotated data, which is oftentimes unavailable or impractical. One way to circumvent this issue is to utilize algorithms that exploit a limited amount of labeled data. This paper aims to review such state-of-the-art algorithms that use a limited number of annotated samples. We explain the fundamental principles of self-supervised learning, generative models, few-shot learning, and semi-supervised learning and summarize their applications in cardiac, abdomen, and brain MRI segmentation. Throughout this review, we highlight algorithms that can be employed based on the quantity of annotated data available. We also present a comprehensive list of notable publicly available MRI segmentation datasets. To conclude, we discuss possible future directions of the field-including emerging algorithms, such as contrastive language-image pretraining, and potential combinations across the methods discussed-that can further increase the efficacy of image segmentation with limited labels.

Transformers and large language models in healthcare: A review

With Artificial Intelligence (AI) increasingly permeating various aspects of society, including healthcare, the adoption of the Transformers neural network architecture is rapidly changing many applications. Transformer is a type of deep learning architecture initially developed to solve general-purpose Natural Language Processing (NLP) tasks and has subsequently been adapted in many fields, including healthcare. In this survey paper, we provide an overview of how this architecture has been adopted to analyze various forms of healthcare data, including clinical NLP, medical imaging, structured Electronic Health Records (EHR), social media, bio-physiological signals, biomolecular sequences. Furthermore, which have also include the articles that used the transformer architecture for generating surgical instructions and predicting adverse outcomes after surgeries under the umbrella of critical care. Under diverse settings, these models have been used for clinical diagnosis, report generation, data reconstruction, and drug/protein synthesis. Finally, we also discuss the benefits and limitations of using transformers in healthcare and examine issues such as computational cost, model interpretability, fairness, alignment with human values, ethical implications, and environmental impact.

nBEST: Deep-learning-based non-human primates Brain Extraction and Segmentation Toolbox across ages, sites and species

Accurate processing and analysis of non-human primate (NHP) brain magnetic resonance imaging (MRI) serves an indispensable role in understanding brain evolution, development, aging, and diseases. Despite the accumulation of diverse NHP brain MRI datasets at various developmental stages and from various imaging sites/scanners, existing computational tools designed for human MRI typically perform poor on NHP data, due to huge differences in brain sizes, morphologies, and imaging appearances across species, sites, and ages, highlighting the imperative for NHP-specialized MRI processing tools. To address this issue, in this paper, we present a robust, generic, and fully automated computational pipeline, called non-human primates Brain Extraction and Segmentation Toolbox (nBEST), whose main functionality includes brain extraction, non-cerebrum removal, and tissue segmentation. Building on cutting-edge deep learning techniques by employing lifelong learning to flexibly integrate data from diverse NHP populations and innovatively constructing 3D U-NeXt architecture, nBEST can well handle structural NHP brain MR images from multi-species, multi-site, and multi-developmental-stage (from neonates to the elderly). We extensively validated nBEST based on, to our knowledge, the largest assemblage dataset in NHP brain studies, encompassing 1,469 scans with 11 species (e.g., rhesus macaques, cynomolgus macaques, chimpanzees, marmosets, squirrel monkeys, etc.) from 23 independent datasets. Compared to alternative tools, nBEST outperforms in precision, applicability, robustness, comprehensiveness, and generalizability, greatly benefiting downstream longitudinal, cross-sectional, and cross-species quantitative analyses. We have made nBEST an open-source toolbox (https://github.com/TaoZhong11/nBEST) and we are committed to its continual refinement through lifelong learning with incoming data to greatly contribute to the research field.

Functional connectome through the human life span

The lifespan growth of the functional connectome remains unknown. Here, we assemble task-free functional and structural magnetic resonance imaging data from 33,250 individuals aged 32 postmenstrual weeks to 80 years from 132 global sites. We report critical inflection points in the nonlinear growth curves of the global mean and variance of the connectome, peaking in the late fourth and late third decades of life, respectively. After constructing a fine-grained, lifespan-wide suite of system-level brain atlases, we show distinct maturation timelines for functional segregation within different systems. Lifespan growth of regional connectivity is organized along a primary-to-association cortical axis. These connectome-based normative models reveal substantial individual heterogeneities in functional brain networks in patients with autism spectrum disorder, major depressive disorder, and Alzheimer's disease. These findings elucidate the lifespan evolution of the functional connectome and can serve as a normative reference for quantifying individual variation in development, aging, and neuropsychiatric disorders.

VINNA for neonates: Orientation independence through latent augmentations

A robust, fast, and accurate segmentation of neonatal brain images is highly desired to better understand and detect changes during development and disease, specifically considering the rise in imaging studies for this cohort. Yet, the limited availability of ground truth datasets, lack of standardized acquisition protocols, and wide variations of head positioning in the scanner pose challenges for method development. A few automated image analysis pipelines exist for newborn brain Magnetic Resonance Image (MRI) segmentation, but they often rely on time-consuming non-linear spatial registration procedures and require resampling to a common resolution, subject to loss of information due to interpolation and down-sampling. Without registration and image resampling, variations with respect to head positions and voxel resolutions have to be addressed differently. In deep learning, external augmentations such as rotation, translation, and scaling are traditionally used to artificially expand the representation of spatial variability, which subsequently increases both the training dataset size and robustness. However, these transformations in the image space still require resampling, reducing accuracy specifically in the context of label interpolation. We recently introduced the concept of resolution-independence with the Voxel-size Independent Neural Network framework, VINN. Here, we extend this concept by additionally shifting all rigid-transforms into the network architecture with a four degree of freedom (4-DOF) transform module, enabling resolution-aware internal augmentations (VINNA) for deep learning. In this work, we show that VINNA (i) significantly outperforms state-of-the-art external augmentation approaches, (ii) effectively addresses the head variations present specifically in newborn datasets, and (iii) retains high segmentation accuracy across a range of resolutions (0.5-1.0 mm). Furthermore, the 4-DOF transform module together with internal augmentations is a powerful, general approach to implement spatial augmentation without requiring image or label interpolation. The specific network application to newborns will be made publicly available as VINNA4neonates.

An end-to-end infant brain parcellation pipeline

Objective

Accurate infant brain parcellation is crucial for understanding early brain development; however, it is challenging due to the inherent low tissue contrast, high noise, and severe partial volume effects in infant magnetic resonance images (MRIs). The aim of this study was to develop an end-to-end pipeline that enabled accurate parcellation of infant brain MRIs.

Methods

We proposed an end-to-end pipeline that employs a two-stage global-to-local approach for accurate parcellation of infant brain MRIs. Specifically, in the global regions of interest (ROIs) localization stage, a combination of transformer and convolution operations was employed to capture both global spatial features and fine texture features, enabling an approximate localization of the ROIs across the whole brain. In the local ROIs refinement stage, leveraging the position priors from the first stage along with the raw MRIs, the boundaries o the ROIs are refined for a more accurate parcellation.

Results

We utilized the Dice ratio to evaluate the accuracy of parcellation results. Results on 263 subjects from National Database for Autism Research (NDAR), Baby Connectome Project (BCP) and Cross-site datasets demonstrated the better accuracy and robustness of our method than other competing methods.

Conclusion

Our end-to-end pipeline may be capable of accurately parcellating 6-month-old infant brain MRIs.

A Novel Mis-Seg-Focus Loss Function Based on a Two-Stage nnU-Net Framework for Accurate Brain Tissue Segmentation

Brain tissue segmentation plays a critical role in the diagnosis, treatment, and study of brain diseases. Accurately identifying these boundaries is essential for improving segmentation accuracy. However, distinguishing boundaries between different brain tissues can be challenging, as they often overlap. Existing deep learning methods primarily calculate the overall segmentation results without adequately addressing local regions, leading to error propagation and mis-segmentation along boundaries. In this study, we propose a novel mis-segmentation-focused loss function based on a two-stage nnU-Net framework. Our approach aims to enhance the model's ability to handle ambiguous boundaries and overlapping anatomical structures, thereby achieving more accurate brain tissue segmentation results. Specifically, the first stage targets the identification of mis-segmentation regions using a global loss function, while the second stage involves defining a mis-segmentation loss function to adaptively adjust the model, thus improving its capability to handle ambiguous boundaries and overlapping anatomical structures. Experimental evaluations on two datasets demonstrate that our proposed method outperforms existing approaches both quantitatively and qualitatively.

Updates to the Melbourne Children's Regional Infant Brain Software Package (M-CRIB-S)

The delineation of cortical areas on magnetic resonance images (MRI) is important for understanding the complexities of the developing human brain. The previous version of the Melbourne Children's Regional Infant Brain (M-CRIB-S) (Adamson et al. Scientific Reports, 10(1), 10, 2020) is a software package that performs whole-brain segmentation, cortical surface extraction and parcellation of the neonatal brain. Available cortical parcellation schemes in the M-CRIB-S are the adult-compatible 34- and 31-region per hemisphere Desikan-Killiany (DK) and Desikan-Killiany-Tourville (DKT), respectively. We present a major update to the software package which achieves two aims: 1) to make the voxel-based segmentation outputs derived from the Freesurfer-compatible M-CRIB scheme, and 2) to improve the accuracy of whole-brain segmentation and cortical surface extraction. Cortical surface extraction has been improved with additional steps to improve penetration of the inner surface into thin gyri. The improved cortical surface extraction is shown to increase the robustness of measures such as surface area, cortical thickness, and cortical volume.

Dual consistency regularization with subjective logic for semi-supervised medical image segmentation

Semi-supervised learning plays a vital role in computer vision tasks, particularly in medical image analysis. It significantly reduces the time and cost involved in labeling data. Current methods primarily focus on consistency regularization and the generation of pseudo labels. However, due to the model's poor awareness of unlabeled data, aforementioned methods may misguide the model. To alleviate this problem, we propose a dual consistency regularization with subjective logic for semi-supervised medical image segmentation. Specifically, we introduce subjective logic into our semi-supervised medical image segmentation task to estimate uncertainty, and based on the consistency hypothesis, we construct dual consistency regularization under weak and strong perturbations to guide the model's learning from unlabeled data. To evaluate the performance of the proposed method, we performed experiments on three widely used datasets: ACDC, LA, and Pancreas. Experiments show that the proposed method achieved improvement compared with other state-of-the-art (SOTA) methods.

Deep learning techniques for isointense infant brain tissue segmentation: a systematic literature review

Introduction

To improve comprehension of initial brain growth in wellness along with sickness, it is essential to precisely segment child brain magnetic resonance imaging (MRI) into white matter (WM) and gray matter (GM), along with cerebrospinal fluid (CSF). Nonetheless, in the isointense phase (6-8 months of age), the inborn myelination and development activities, WM along with GM display alike stages of intensity in both T1-weighted and T2-weighted MRI, making tissue segmentation extremely difficult.

Methods

The comprehensive review of studies related to isointense brain MRI segmentation approaches is highlighted in this publication. The main aim and contribution of this study is to aid researchers by providing a thorough review to make their search for isointense brain MRI segmentation easier. The systematic literature review is performed from four points of reference: (1) review of studies concerning isointense brain MRI segmentation; (2) research contribution and future works and limitations; (3) frequently applied evaluation metrics and datasets; (4) findings of this studies.

Results and discussion

The systemic review is performed on studies that were published in the period of 2012 to 2022. A total of 19 primary studies of isointense brain MRI segmentation were selected to report the research question stated in this review.

Recent progress in transformer-based medical image analysis

The transformer is primarily used in the field of natural language processing. Recently, it has been adopted and shows promise in the computer vision (CV) field. Medical image analysis (MIA), as a critical branch of CV, also greatly benefits from this state-of-the-art technique. In this review, we first recap the core component of the transformer, the attention mechanism, and the detailed structures of the transformer. After that, we depict the recent progress of the transformer in the field of MIA. We organize the applications in a sequence of different tasks, including classification, segmentation, captioning, registration, detection, enhancement, localization, and synthesis. The mainstream classification and segmentation tasks are further divided into eleven medical image modalities. A large number of experiments studied in this review illustrate that the transformer-based method outperforms existing methods through comparisons with multiple evaluation metrics. Finally, we discuss the open challenges and future opportunities in this field. This task-modality review with the latest contents, detailed information, and comprehensive comparison may greatly benefit the broad MIA community.

Fine-grained functional parcellation maps of the infant cerebral cortex

Resting-state functional MRI (rs-fMRI) is widely used to examine the dynamic brain functional development of infants, but these studies typically require precise cortical parcellation maps, which cannot be directly borrowed from adult-based functional parcellation maps due to the substantial differences in functional brain organization between infants and adults. Creating infant-specific cortical parcellation maps is thus highly desired but remains challenging due to difficulties in acquiring and processing infant brain MRIs. In this study, we leveraged 1064 high-resolution longitudinal rs-fMRIs from 197 typically developing infants and toddlers from birth to 24 months who participated in the Baby Connectome Project to develop the first set of infant-specific, fine-grained, surface-based cortical functional parcellation maps. To establish meaningful cortical functional correspondence across individuals, we performed cortical co-registration using both the cortical folding geometric features and the local gradient of functional connectivity (FC). Then we generated both age-related and age-independent cortical parcellation maps with over 800 fine-grained parcels during infancy based on aligned and averaged local gradient maps of FC across individuals. These parcellation maps reveal complex functional developmental patterns, such as changes in local gradient, network size, and local efficiency, especially during the first 9 postnatal months. Our generated fine-grained infant cortical functional parcellation maps are publicly available at https://www.nitrc.org/projects/infantsurfatlas/ for advancing the pediatric neuroimaging field.

WRANet: wavelet integrated residual attention U-Net network for medical image segmentation

Medical image segmentation is crucial for the diagnosis and analysis of disease. Deep convolutional neural network methods have achieved great success in medical image segmentation. However, they are highly susceptible to noise interference during the propagation of the network, where weak noise can dramatically alter the network output. As the network deepens, it can face problems such as gradient explosion and vanishing. To improve the robustness and segmentation performance of the network, we propose a wavelet residual attention network (WRANet) for medical image segmentation. We replace the standard downsampling modules (e.g., maximum pooling and average pooling) in CNNs with discrete wavelet transform, decompose the features into low- and high-frequency components, and remove the high-frequency components to eliminate noise. At the same time, the problem of feature loss can be effectively addressed by introducing an attention mechanism. The combined experimental results show that our method can effectively perform aneurysm segmentation, achieving a Dice score of 78.99%, an IoU score of 68.96%, a precision of 85.21%, and a sensitivity score of 80.98%. In polyp segmentation, a Dice score of 88.89%, an IoU score of 81.74%, a precision rate of 91.32%, and a sensitivity score of 91.07% were achieved. Furthermore, our comparison with state-of-the-art techniques demonstrates the competitiveness of the WRANet network.

Minimizing Estimated Risks on Unlabeled Data: A New Formulation for Semi-Supervised Medical Image Segmentation

Supervised segmentation can be costly, particularly in applications of biomedical image analysis where large scale manual annotations from experts are generally too expensive to be available. Semi-supervised segmentation, able to learn from both the labeled and unlabeled images, could be an efficient and effective alternative for such scenarios. In this work, we propose a new formulation based on risk minimization, which makes full use of the unlabeled images. Different from most of the existing approaches which solely explicitly guarantee the minimization of prediction risks from the labeled training images, the new formulation also considers the risks on unlabeled images. Particularly, this is achieved via an unbiased estimator, based on which we develop a general framework for semi-supervised image segmentation. We validate this framework on three medical image segmentation tasks, namely cardiac segmentation on ACDC2017, optic cup and disc segmentation on REFUGE dataset and 3D whole heart segmentation on MM-WHS dataset. Results show that the proposed estimator is effective, and the segmentation method achieves superior performance and demonstrates great potential compared to the other state-of-the-art approaches. Our code and data will be released via https://zmiclab.github.io/projects.html, once the manuscript is accepted for publication.

An attention-based context-informed deep framework for infant brain subcortical segmentation

Precise segmentation of subcortical structures from infant brain magnetic resonance (MR) images plays an essential role in studying early subcortical structural and functional developmental patterns and diagnosis of related brain disorders. However, due to the dynamic appearance changes, low tissue contrast, and tiny subcortical size in infant brain MR images, infant subcortical segmentation is a challenging task. In this paper, we propose a context-guided, attention-based, coarse-to-fine deep framework to precisely segment the infant subcortical structures. At the coarse stage, we aim to directly predict the signed distance maps (SDMs) from multi-modal intensity images, including T1w, T2w, and the ratio of T1w and T2w images, with an SDM-Unet, which can leverage the spatial context information, including the structural position information and the shape information of the target structure, to generate high-quality SDMs. At the fine stage, the predicted SDMs, which encode spatial-context information of each subcortical structure, are integrated with the multi-modal intensity images as the input to a multi-source and multi-path attention Unet (M2A-Unet) for achieving refined segmentation. Both the 3D spatial and channel attention blocks are added to guide the M2A-Unet to focus more on the important subregions and channels. We additionally incorporate the inner and outer subcortical boundaries as extra labels to help precisely estimate the ambiguous boundaries. We validate our method on an infant MR image dataset and on an unrelated neonatal MR image dataset. Compared to eleven state-of-the-art methods, the proposed framework consistently achieves higher segmentation accuracy in both qualitative and quantitative evaluations of infant MR images and also exhibits good generalizability in the neonatal dataset.

MAGNET: A MODALITY-AGNOSTIC NETWORK FOR 3D MEDICAL IMAGE SEGMENTATION

In this paper, we proposed MAGNET, a novel modality-agnostic network for 3D medical image segmentation. Different from existing learning methods, MAGNET is specifically designed to handle real medical situations where multiple modalities/sequences are available during model training, but fewer ones are available or used at time of clinical practice. Our results on multiple datasets show that MAGNET trained on multi-modality data has the unique ability to perform predictions using any subset of training imaging modalities. It outperforms individually trained uni-modality models while can further boost performance when more modalities are available at testing.

Transforming medical imaging with Transformers? A comparative review of key properties, current progresses, and future perspectives

Transformer, one of the latest technological advances of deep learning, has gained prevalence in natural language processing or computer vision. Since medical imaging bear some resemblance to computer vision, it is natural to inquire about the status quo of Transformers in medical imaging and ask the question: can the Transformer models transform medical imaging? In this paper, we attempt to make a response to the inquiry. After a brief introduction of the fundamentals of Transformers, especially in comparison with convolutional neural networks (CNNs), and highlighting key defining properties that characterize the Transformers, we offer a comprehensive review of the state-of-the-art Transformer-based approaches for medical imaging and exhibit current research progresses made in the areas of medical image segmentation, recognition, detection, registration, reconstruction, enhancement, etc. In particular, what distinguishes our review lies in its organization based on the Transformer's key defining properties, which are mostly derived from comparing the Transformer and CNN, and its type of architecture, which specifies the manner in which the Transformer and CNN are combined, all helping the readers to best understand the rationale behind the reviewed approaches. We conclude with discussions of future perspectives.

MediNet: transfer learning approach with MediNet medical visual database

The rapid development of machine learning has increased interest in the use of deep learning methods in medical research. Deep learning in the medical field is used in disease detection and classification problems in the clinical decision-making process. Large amounts of labeled datasets are often required to train deep neural networks; however, in the medical field, the lack of a sufficient number of images in datasets and the difficulties encountered during data collection are among the main problems. In this study, we propose MediNet, a new 10-class visual dataset consisting of Rontgen (X-ray), Computed Tomography (CT), Magnetic Resonance Imaging (MRI), Ultrasound, and Histopathological images such as calcaneal normal, calcaneal tumor, colon benign colon adenocarcinoma, brain normal, brain tumor, breast benign, breast malignant, chest normal, chest pneumonia. AlexNet, VGG19-BN, Inception V3, DenseNet 121, ResNet 101, EfficientNet B0, Nested-LSTM + CNN, and proposed RdiNet deep learning algorithms are used in the transfer learning for pre-training and classification application. Transfer learning aims to apply previously learned knowledge in a new task. Seven algorithms were trained with the MediNet dataset, and the models obtained from these algorithms, namely feature vectors, were recorded. Pre-training models were used for classification studies on chest X-ray images, diabetic retinopathy, and Covid-19 datasets with the transfer learning technique. In performance measurement, an accuracy of 94.84% was obtained in the traditional classification study for the InceptionV3 model in the classification study performed on the Chest X-Ray Images dataset, and the accuracy was increased 98.71% after the transfer learning technique was applied. In the Covid-19 dataset, the classification success of the DenseNet121 model before pre-trained was 88%, while the performance after the transfer application with MediNet was 92%. In the Diabetic retinopathy dataset, the classification success of the Nested-LSTM + CNN model before pre-trained was 79.35%, while the classification success was 81.52% after the transfer application with MediNet. The comparison of results obtained from experimental studies observed that the proposed method produced more successful results.

Graphical abstract

iBEAT V2.0: a multisite-applicable, deep learning-based pipeline for infant cerebral cortical surface reconstruction

The human cerebral cortex undergoes dramatic and critical development during early postnatal stages. Benefiting from advances in neuroimaging, many infant brain magnetic resonance imaging (MRI) datasets have been collected from multiple imaging sites with different scanners and imaging protocols for the investigation of normal and abnormal early brain development. However, it is extremely challenging to precisely process and quantify infant brain development with these multisite imaging data because infant brain MRI scans exhibit (a) extremely low and dynamic tissue contrast caused by ongoing myelination and maturation and (b) inter-site data heterogeneity resulting from the use of diverse imaging protocols/scanners. Consequently, existing computational tools and pipelines typically perform poorly on infant MRI data. To address these challenges, we propose a robust, multisite-applicable, infant-tailored computational pipeline that leverages powerful deep learning techniques. The main functionality of the proposed pipeline includes preprocessing, brain skull stripping, tissue segmentation, topology correction, cortical surface reconstruction and measurement. Our pipeline can handle both T1w and T2w structural infant brain MR images well in a wide age range (from birth to 6 years of age) and is effective for different imaging protocols/scanners, despite being trained only on the data from the Baby Connectome Project. Extensive comparisons with existing methods on multisite, multimodal and multi-age datasets demonstrate superior effectiveness, accuracy and robustness of our pipeline. We have maintained a website, iBEAT Cloud, for users to process their images with our pipeline ( http://www.ibeat.cloud ), which has successfully processed over 16,000 infant MRI scans from more than 100 institutions with various imaging protocols/scanners.

3D-MASNet: 3D mixed-scale asymmetric convolutional segmentation network for 6-month-old infant brain MR images

Precise segmentation of infant brain magnetic resonance (MR) images into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) are essential for studying neuroanatomical hallmarks of early brain development. However, for 6-month-old infants, the extremely low-intensity contrast caused by inherent myelination hinders accurate tissue segmentation. Existing convolutional neural networks (CNNs) based segmentation models for this task generally employ single-scale symmetric convolutions, which are inefficient for encoding the isointense tissue boundaries in baby brain images. Here, we propose a 3D mixed-scale asymmetric convolutional segmentation network (3D-MASNet) framework for brain MR images of 6-month-old infants. We replaced the traditional convolutional layer of an existing to-be-trained network with a 3D mixed-scale convolution block consisting of asymmetric kernels (MixACB) during the training phase and then equivalently converted it into the original network. Five canonical CNN segmentation models were evaluated using both T1- and T2-weighted images of 23 6-month-old infants from iSeg-2019 datasets, which contained manual labels as ground truth. MixACB significantly enhanced the average accuracy of all five models and obtained the most considerable improvement in the fully convolutional network model (CC-3D-FCN) and the highest performance in the Dense U-Net model. This approach further obtained Dice coefficient accuracies of 0.931, 0.912, and 0.961 in GM, WM, and CSF, respectively, ranking first among 30 teams on the validation dataset of the iSeg-2019 Grand Challenge. Thus, the proposed 3D-MASNet can improve the accuracy of existing CNNs-based segmentation models as a plug-and-play solution that offers a promising technique for future infant brain MRI studies.

ConvNets for automatic detection of polyglutamine SCAs from brain MRIs: state of the art applications

Polyglutamine spinocerebellar ataxias (polyQ SCAs) are a group of neurodegenerative diseases, clinically and genetically heterogeneous, characterized by loss of balance and motor coordination due to dysfunction of the cerebellum and its connections. The diagnosis of each type of polyQ SCA, alongside with genetic tests, includes medical images analysis, and its automation may help specialists to distinguish between each type. Convolutional neural networks (ConvNets or CNNs) have been recently used for medical image processing, with outstanding results. In this work, we present the main clinical and imaging features of polyglutamine SCAs, and the basics of CNNs. Finally, we review studies that have used this approach to automatically process brain medical images and may be applied to SCAs detection. We conclude by discussing the possible limitations and opportunities of using ConvNets for SCAs diagnose in the future.

Deep learning in bladder cancer imaging: A review

Deep learning (DL) is a rapidly developing field in machine learning (ML). The concept of deep learning originates from research on artificial neural networks and is an upgrade of traditional neural networks. It has achieved great success in various domains and has shown potential in solving medical problems, particularly when using medical images. Bladder cancer (BCa) is the tenth most common cancer in the world. Imaging, as a safe, noninvasive, and relatively inexpensive technique, is a powerful tool to aid in the diagnosis and treatment of bladder cancer. In this review, we provide an overview of the latest progress in the application of deep learning to the imaging assessment of bladder cancer. First, we review the current deep learning approaches used for bladder segmentation. We then provide examples of how deep learning helps in the diagnosis, staging, and treatment management of bladder cancer using medical images. Finally, we summarize the current limitations of deep learning and provide suggestions for future improvements.

A bidirectional multilayer contrastive adaptation network with anatomical structure preservation for unpaired cross-modality medical image segmentation

Multi-modal medical image segmentation has achieved great success through supervised deep learning networks. However, because of domain shift and limited annotation information, unpaired cross-modality segmentation tasks are still challenging. The unsupervised domain adaptation (UDA) methods can alleviate the segmentation degradation of cross-modality segmentation by knowledge transfer between different domains, but current methods still suffer from the problems of model collapse, adversarial training instability, and mismatch of anatomical structures. To tackle these issues, we propose a bidirectional multilayer contrastive adaptation network (BMCAN) for unpaired cross-modality segmentation. The shared encoder is first adopted for learning modality-invariant encoding representations in image synthesis and segmentation simultaneously. Secondly, to retain the anatomical structure consistency in cross-modality image synthesis, we present a structure-constrained cross-modality image translation approach for image alignment. Thirdly, we construct a bidirectional multilayer contrastive learning approach to preserve the anatomical structures and enhance encoding representations, which utilizes two groups of domain-specific multilayer perceptron (MLP) networks to learn modality-specific features. Finally, a semantic information adversarial learning approach is designed to learn structural similarities of semantic outputs for output space alignment. Our proposed method was tested on three different cross-modality segmentation tasks: brain tissue, brain tumor, and cardiac substructure segmentation. Compared with other UDA methods, experimental results show that our proposed BMCAN achieves state-of-the-art segmentation performance on the above three tasks, and it has fewer training components and better feature representations for overcoming overfitting and domain shift problems. Our proposed method can efficiently reduce the annotation burden of radiologists in cross-modality image analysis.

Learning to detect boundary information for brain image segmentation

MRI brain images are always of low contrast, which makes it difficult to identify to which area the information at the boundary of brain images belongs. This can make the extraction of features at the boundary more challenging, since those features can be misleading as they might mix properties of different brain regions. Hence, to alleviate such a problem, image boundary detection plays a vital role in medical image segmentation, and brain segmentation in particular, as unclear boundaries can worsen brain segmentation results. Yet, given the low quality of brain images, boundary detection in the context of brain image segmentation remains challenging. Despite the research invested to improve boundary detection and brain segmentation, these two problems were addressed independently, i.e., little attention was paid to applying boundary detection to brain segmentation tasks. Therefore, in this paper, we propose a boundary detection-based model for brain image segmentation. To this end, we first design a boundary segmentation network for detecting and segmenting images brain tissues. Then, we design a boundary information module (BIM) to distinguish boundaries from the three different brain tissues. After that, we add a boundary attention gate (BAG) to the encoder output layers of our transformer to capture more informative local details. We evaluate our proposed model on two datasets of brain tissue images, including infant and adult brains. The extensive evaluation experiments of our model show better performance (a Dice Coefficient (DC) accuracy of up to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$5.3\%$$\end{document}5.3% compared to the state-of-the-art models) in detecting and segmenting brain tissue images.

Recurrent Tissue-Aware Network for Deformable Registration of Infant Brain MR Images

Deformable registration is fundamental to longitudinal and population-based image analyses. However, it is challenging to precisely align longitudinal infant brain MR images of the same subject, as well as cross-sectional infant brain MR images of different subjects, due to fast brain development during infancy. In this paper, we propose a recurrently usable deep neural network for the registration of infant brain MR images. There are three main highlights of our proposed method. (i) We use brain tissue segmentation maps for registration, instead of intensity images, to tackle the issue of rapid contrast changes of brain tissues during the first year of life. (ii) A single registration network is trained in a one-shot manner, and then recurrently applied in inference for multiple times, such that the complex deformation field can be recovered incrementally. (iii) We also propose both the adaptive smoothing layer and the tissue-aware anti-folding constraint into the registration network to ensure the physiological plausibility of estimated deformations without degrading the registration accuracy. Experimental results, in comparison to the state-of-the-art registration methods, indicate that our proposed method achieves the highest registration accuracy while still preserving the smoothness of the deformation field. The implementation of our proposed registration network is available online https://github.com/Barnonewdm/ACTA-Reg-Net.

A cascaded nested network for 3T brain MR image segmentation guided by 7T labeling

Accurate segmentation of the brain into gray matter, white matter, and cerebrospinal fluid using magnetic resonance (MR) imaging is critical for visualization and quantification of brain anatomy. Compared to 3T MR images, 7T MR images exhibit higher tissue contrast that is contributive to accurate tissue delineation for training segmentation models. In this paper, we propose a cascaded nested network (CaNes-Net) for segmentation of 3T brain MR images, trained by tissue labels delineated from the corresponding 7T images. We first train a nested network (Nes-Net) for a rough segmentation. The second Nes-Net uses tissue-specific geodesic distance maps as contextual information to refine the segmentation. This process is iterated to build CaNes-Net with a cascade of Nes-Net modules to gradually refine the segmentation. To alleviate the misalignment between 3T and corresponding 7T MR images, we incorporate a correlation coefficient map to allow well-aligned voxels to play a more important role in supervising the training process. We compared CaNes-Net with SPM and FSL tools, as well as four deep learning models on 18 adult subjects and the ADNI dataset. Our results indicate that CaNes-Net reduces segmentation errors caused by the misalignment and improves segmentation accuracy substantially over the competing methods.

Automatic brain segmentation in preterm infants with post-hemorrhagic hydrocephalus using 3D Bayesian U-Net

Post‐hemorrhagic hydrocephalus (PHH) is a severe complication of intraventricular hemorrhage (IVH) in very preterm infants. PHH monitoring and treatment decisions rely heavily on manual and subjective two‐dimensional measurements of the ventricles. Automatic and reliable three‐dimensional (3D) measurements of the ventricles may provide a more accurate assessment of PHH, and lead to improved monitoring and treatment decisions. To accurately and efficiently obtain these 3D measurements, automatic segmentation of the ventricles can be explored. However, this segmentation is challenging due to the large ventricular anatomical shape variability in preterm infants diagnosed with PHH. This study aims to (a) propose a Bayesian U‐Net method using 3D spatial concrete dropout for automatic brain segmentation (with uncertainty assessment) of preterm infants with PHH; and (b) compare the Bayesian method to three reference methods: DenseNet, U‐Net, and ensemble learning using DenseNets and U‐Nets. A total of 41 T₂‐weighted MRIs from 27 preterm infants were manually segmented into lateral ventricles, external CSF, white and cortical gray matter, brainstem, and cerebellum. These segmentations were used as ground truth for model evaluation. All methods were trained and evaluated using 4‐fold cross‐validation and segmentation endpoints, with additional uncertainty endpoints for the Bayesian method. In the lateral ventricles, segmentation endpoint values for the DenseNet, U‐Net, ensemble learning, and Bayesian U‐Net methods were mean Dice score = 0.814 ± 0.213, 0.944 ± 0.041, 0.942 ± 0.042, and 0.948 ± 0.034 respectively. Uncertainty endpoint values for the Bayesian U‐Net were mean recall = 0.953 ± 0.037, mean  negative predictive value = 0.998 ± 0.005, mean accuracy = 0.906 ± 0.032, and mean AUC = 0.949 ± 0.031. To conclude, the Bayesian U‐Net showed the best segmentation results across all methods and provided accurate uncertainty maps. This method may be used in clinical practice for automatic brain segmentation of preterm infants with PHH, and lead to better PHH monitoring and more informed treatment decisions.

Unified framework for early stage status prediction of autism based on infant structural magnetic resonance imaging

Autism, or autism spectrum disorder (ASD), is a developmental disability that is diagnosed at about 2 years of age based on abnormal behaviors. Existing neuroimaging‐based methods for the prediction of ASD typically focus on functional magnetic resonance imaging (fMRI); however, most of these fMRI‐based studies include subjects older than 5 years of age. Due to challenges in the application of fMRI for infants, structural magnetic resonance imaging (sMRI) has increasingly received attention in the field for early status prediction of ASD. In this study, we propose an automated prediction framework based on infant sMRI at about 24 months of age. Specifically, by leveraging an infant‐dedicated pipeline, iBEAT V2.0 Cloud, we derived segmentation and parcellation maps from infant sMRI. We employed a convolutional neural network to extract features from pairwise maps and a Siamese network to distinguish whether paired subjects were from the same or different classes. As compared to T1w imaging without segmentation and parcellation maps, our proposed approach with segmentation and parcellation maps yielded greater sensitivity, specificity, and accuracy of ASD prediction, which was validated using two datasets with different imaging protocols/scanners and was confirmed by receiver operating characteristic analysis. Furthermore, comparison with state‐of‐the‐art methods demonstrated the superior effectiveness and robustness of the proposed method. Finally, attention maps were generated to identify subject‐specific autism effects, supporting the reasonability of the predictive results. Collectively, these findings demonstrate the utility of our unified framework for the early‐stage status prediction of ASD by sMRI.

HybridCTrm: Bridging CNN and Transformer for Multimodal Brain Image Segmentation

Multimodal medical image segmentation is always a critical problem in medical image segmentation. Traditional deep learning methods utilize fully CNNs for encoding given images, thus leading to deficiency of long-range dependencies and bad generalization performance. Recently, a sequence of Transformer-based methodologies emerges in the field of image processing, which brings great generalization and performance in various tasks. On the other hand, traditional CNNs have their own advantages, such as rapid convergence and local representations. Therefore, we analyze a hybrid multimodal segmentation method based on Transformers and CNNs and propose a novel architecture, HybridCTrm network. We conduct experiments using HybridCTrm on two benchmark datasets and compare with HyperDenseNet, a network based on fully CNNs. Results show that our HybridCTrm outperforms HyperDenseNet on most of the evaluation metrics. Furthermore, we analyze the influence of the depth of Transformer on the performance. Besides, we visualize the results and carefully explore how our hybrid methods improve on segmentations.

Annotation-efficient deep learning for automatic medical image segmentation

Automatic medical image segmentation plays a critical role in scientific research and medical care. Existing high-performance deep learning methods typically rely on large training datasets with high-quality manual annotations, which are difficult to obtain in many clinical applications. Here, we introduce Annotation-effIcient Deep lEarning (AIDE), an open-source framework to handle imperfect training datasets. Methodological analyses and empirical evaluations are conducted, and we demonstrate that AIDE surpasses conventional fully-supervised models by presenting better performance on open datasets possessing scarce or noisy annotations. We further test AIDE in a real-life case study for breast tumor segmentation. Three datasets containing 11,852 breast images from three medical centers are employed, and AIDE, utilizing 10% training annotations, consistently produces segmentation maps comparable to those generated by fully-supervised counterparts or provided by independent radiologists. The 10-fold enhanced efficiency in utilizing expert labels has the potential to promote a wide range of biomedical applications.

Realistic image normalization for multi-Domain segmentation

Image normalization is a building block in medical image analysis. Conventional approaches are customarily employed on a per-dataset basis. This strategy, however, prevents the current normalization algorithms from fully exploiting the complex joint information available across multiple datasets. Consequently, ignoring such joint information has a direct impact on the processing of segmentation algorithms. This paper proposes to revisit the conventional image normalization approach by, instead, learning a common normalizing function across multiple datasets. Jointly normalizing multiple datasets is shown to yield consistent normalized images as well as an improved image segmentation when intensity shifts are large. To do so, a fully automated adversarial and task-driven normalization approach is employed as it facilitates the training of realistic and interpretable images while keeping performance on par with the state-of-the-art. The adversarial training of our network aims at finding the optimal transfer function to improve both, jointly, the segmentation accuracy and the generation of realistic images. We have evaluated the performance of our normalizer on both infant and adult brain images from the iSEG, MRBrainS and ABIDE datasets. The results indicate that our contribution does provide an improved realism to the normalized images, while retaining a segmentation accuracy at par with the state-of-the-art learnable normalization approaches.

Multi-Scale Self-Supervised Learning for Multi-Site Pediatric Brain MR Image Segmentation with Motion/Gibbs Artifacts

Accurate tissue segmentation of large-scale pediatric brain MR images from multiple sites is essential to characterize early brain development. Due to imaging motion/Gibbs artifacts and multi-site issue (or domain shift issue), it remains a challenge to accurately segment brain tissues from multi-site pediatric MR images. In this paper, we present a multi-scale self-supervised learning (M-SSL) framework to accurately segment tissues for multi-site pediatric brain MR images with artifacts. Specifically, we first work on the downsampled images to estimate coarse tissue probabilities and build a global anatomic guidance. We then train another segmentation model based on the original images to estimate fine tissue probabilities, which are further integrated with the global anatomic guidance to refine the segmentation results. In the testing stage, to alleviate the multi-site issue, we propose an iterative self-supervised learning strategy to train a site-specific segmentation model based on a set of reliable training samples automatically generated for a to-be-segmented site. The experimental results on pediatric brain MR images with real artifacts and multi-site subjects from the iSeg2019 challenge demonstrate that our M-SSL method achieves better performance compared with several state-of-the-art methods.

Narrow Band Active Contour Attention Model for Medical Segmentation

Medical image segmentation is one of the most challenging tasks in medical image analysis and widely developed for many clinical applications. While deep learning-based approaches have achieved impressive performance in semantic segmentation, they are limited to pixel-wise settings with imbalanced-class data problems and weak boundary object segmentation in medical images. In this paper, we tackle those limitations by developing a new two-branch deep network architecture which takes both higher level features and lower level features into account. The first branch extracts higher level feature as region information by a common encoder-decoder network structure such as Unet and FCN, whereas the second branch focuses on lower level features as support information around the boundary and processes in parallel to the first branch. Our key contribution is the second branch named Narrow Band Active Contour (NB-AC) attention model which treats the object contour as a hyperplane and all data inside a narrow band as support information that influences the position and orientation of the hyperplane. Our proposed NB-AC attention model incorporates the contour length with the region energy involving a fixed-width band around the curve or surface. The proposed network loss contains two fitting terms: (i) a high level feature (i.e., region) fitting term from the first branch; (ii) a lower level feature (i.e., contour) fitting term from the second branch including the (ii1) length of the object contour and (ii2) regional energy functional formed by the homogeneity criterion of both the inner band and outer band neighboring the evolving curve or surface. The proposed NB-AC loss can be incorporated into both 2D and 3D deep network architectures. The proposed network has been evaluated on different challenging medical image datasets, including DRIVE, iSeg17, MRBrainS18 and Brats18. The experimental results have shown that the proposed NB-AC loss outperforms other mainstream loss functions: Cross Entropy, Dice, Focal on two common segmentation frameworks Unet and FCN. Our 3D network which is built upon the proposed NB-AC loss and 3DUnet framework achieved state-of-the-art results on multiple volumetric datasets.

Rubik-Net: Learning Spatial Information via Rotation-Driven Convolutions for Brain Segmentation

The accurate segmentation of brain tissue in Magnetic Resonance Image (MRI) slices is essential for assessing neurological conditions and brain diseases. However, it is challenging to segment MRI slices because of the low contrast between different brain tissues and the partial volume effect. A 2-Dimensional (2-D) convolutional network cannot handle such volumetric medical image data well because it overlooks spatial information between MRI slices. Although 3-Dimensional (3-D) convolutions capture volumetric spatial information, they have not been fully exploited to enhance the representative ability of deep networks; moreover, they may lead to overfitting in the case of insufficient training data. In this paper, we propose a novel convolutional mechanism, termed Rubik convolution, to capture multi dimensional information between MRI slices. Rubik convolution rotates the axis of a set of consecutive slices, which enables a 2-D convolution kernel to extract features of each axial plane simultaneously. Next, feature maps are rotated back to fuse multidimensional information by the Max-View-Maps. Furthermore, we propose an efficient 2-D convolutional network, namely Rubik-Net, where the residual connections and the bottleneck structure are used to enhance information transmission and reduce the number of network parameters. The proposed Rubik-Net shows promising results on iSeg2017, iSeg2019, IBSR and Brainweb datasets in terms of segmentation accuracy. In particular, we achieved the best results in 95th-percentile Hausdorff distance and average surface distance in cerebrospinal fluid segmentation on the most challenging iSeg2019 dataset. The experiments indicate that Rubik-Net improves the accuracy and efficiency of medical image segmentation. Moreover, Rubik convolution can be easily embedded into existing 2-D convolutional networks.